Results from the first ARM diffuse horizontal shortwave irradiance comparison

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D3, 4108, doi: /2002jd002825, 2003 Results from the first ARM diffuse horizontal shortwave irradiance comparison J. J. Michalsky, 1 R. Dolce, 2 E. G. Dutton, 3 M. Haeffelin, 4 G. Major, 5 J. A. Schlemmer, 1 D. W. Slater, 6 J. R. Hickey, 7 W. Q. Jeffries, 8 A. Los, 2 D. Mathias, 9 L. J. B. McArthur, 10 R. Philipona, 11 I. Reda, 12 and T. Stoffel 12 Received 5 August 2002; revised 6 November 2002; accepted 6 December 2002; published 7 February [1] The first intensive observation period (IOP) dedicated exclusively to the measurement of diffuse horizontal shortwave irradiance was held in the Fall 2001 at the central facility of the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site with the cooperation of the Baseline Surface Radiation Network (BSRN) community. The purpose of the study was to compare diffuse irradiance measurements among most commercial pyranometers and a few prototypes calibrated independently using current practices. The hope was to achieve a consensus for this measurement with the goal of improving the uncertainty of shortwave diffuse irradiance measurements. All diffuse broadband measurements were made using the same type of two-axis tracker with the direct beam blocked by shading balls. Tracking was excellent during the IOP with no lost data associated with tracker problems. Fourteen simultaneous measurements were obtained over a two-week period under mostly clear skies with low to moderate aerosol loading. Totally overcast data were obtained during the morning of one day. Five of the measurements are reproducible to about 2 W/m 2 at the 95% confidence level. Three more agree with the mean of these five to about 4 W/m 2 at the 95% confidence level after correction for thermal offsets. INDEX TERMS: 0305 Atmospheric Composition and Structure: Aerosols and particles (0345, 4801); 0394 Atmospheric Composition and Structure: Instruments and techniques; 0360 Atmospheric Composition and Structure: Transmission and scattering of radiation; KEYWORDS: pyranometer, diffuse horizontal irradiance, zero irradiance offset, penumbral corrections, reproducibility, intensive observation period Citation: Michalsky, J. J., et al., Results from the first ARM diffuse horizontal shortwave irradiance comparison, J. Geophys. Res., 108(D3), 4108, doi: /2002jd002825, Introduction 1 Atmospheric Sciences Research Center, State University of New York, Albany, New York, USA. 2 Kipp & Zonen, Inc., Bohemia, New York, USA and Delft, Netherlands. 3 Climate Monitoring and Diagnostics Laboratory, National Oceanic and Atmospheric Administration, Boulder, Colorado, USA. 4 Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA. 5 Budapest University of Economic Sciences and Public Administration, Budapest, Hungary. 6 Pacific Northwest National Laboratory, Richland, Washington, USA. 7 The Eppley Laboratory, Inc., Newport, Rhode Island, USA. 8 Yankee Environmental Systems, Inc., Turners Falls, Massachusetts, USA. 9 Carter-Scott Design, Brunswick, Victoria, Australia. 10 Meteorological Service of Canada, Downsview, Ontario, Canada. 11 Physikalisch-Meteorologisches Observatorium and World Radiation Center, Davos, Switzerland. 12 National Renewable Energy Laboratory, Golden, Colorado, USA. Copyright 2003 by the American Geophysical Union /03/2002JD [2] Shortwave irradiance measurements have improved significantly in the past few years as scientists have pushed for better measurements to rigorously test models of radiative transfer. Models and measurements of the direct normal irradiance have been shown to agree to within measurement and modeling errors [Kato et al., 1997; Halthore et al., 1997]. Since the direct is measured with a small uncertainty using an absolute reference instrument (the self-calibrating cavity radiometer), it is felt that the model and measurement agreement gives us a high level of certainty that the extinction inputs to the model, including water vapor, ozone, and aerosol, are reasonable. When model inputs that satisfy direct irradiance closure were used to calculate diffuse horizontal irradiance, Halthore and Schwartz [2000] found that model diffuse irradiances exceeded measurements for clean, clear-sky conditions. Zender et al. [1997] and Valero and Bush [1999] compared measured total horizontal irradiance and models and were able to reach agreement by assuming modest absorption by the aerosol. [3] The discussion above suggests that we can achieve closure in direct beam comparisons, but closure for diffuse modeling and measurements depends on the assumed aerosol properties, surface reflectance, and measurements of diffuse irradiance. This suggests that there is a need to concentrate our efforts on better measurements of aerosol optical properties, not just optical depth, and AAC 9-1

2 AAC 9-2 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Table 1. Comparison Instruments, Ventilation Type, and Calibration Method Instrument Symbol Ventilation Calibration Technique Carter-Scott Design EQ08-A (prototype) eq08 None Australian BoM a Eppley PSP case+dome temps psp-mh Moderate volume Forgan b CIMEL black&white cimel None Forgan b EKO MS801 eko None CMDL c Kipp&Zonen CM11 cm11 High volume Kipp&Zonen d Kipp&Zonen CM22 cm22 High volume Kipp&Zonen d Kipp&Zonen CM22 modified cm22-rp High volume plus heating WRC e Eppley Moderate volume Eppley dome f Eppley (prototype) new_epp_bw Moderate volume Eppley outside g Kipp&Zonen CM21 cm21 Moderate volume MSC f Schenk Star schenk None MSC f Eppley PSP psp Moderate volume Eppley dome f YES, Inc. Isothermal Pyranometer (prototype) yes Moderate volume YES h Scripps (prototype) tsbr None SIO i a Bureau of Meteorology; component sum using Eppley HF for direct plus Kipp & Zonen CM11 shaded compared at 45 sun position. b Forgan method explained by Forgan [1996]. c Component sum using Eppley HF for direct plus offset correct Eppley PSP and/or 8 48 for diffuse. d K&Z use 1000 W tungsten-halogen lamp over reference and test pyranometers; reference calibrated with standard WRC method. e Component sum using world standard group (WSG) for direct plus shaded PSP is standard WRC method; cm22-rp used shade/unshade compared to WSG. f Dome methods use diffuse reflection from inside of white dome that are illuminated by incandescent lights; no direct illumination of sensors by incandescent lamps. Test instrument compared to calibrated reference. Calibrated reference by component sum or shade/unshade. Reference pyranometer is of same type. g Shade/unshade compared to Eppley HF and component sum. h Outdoor side-by-side-comparison to Eppley PSP. i SIO instrument fitted with 5 field-limiting aperture and compared to Eppley HF. surface albedo and on improving the accuracy of measurements of diffuse irradiance, which is the subject of this paper. [4] Diffuse horizontal shortwave irradiance sensors are pyranometers that are mounted on solar trackers and have direct normal solar irradiance blocked using a tracking ball or disk. In this study, 14 pyranometers were arranged in this way to acquire simultaneous 1-min averaged diffuse measurements over a two-week period. Each pyranometer had a unique design or unique operational mode. It was a critical point that we used only pyranometers with an independent calibration by the instrument provider. The instruments calibrations were obtained using several methods. [5] In the component-sum technique, pyranometers are calibrated in full sun by comparing them to a reference that measures direct and diffuse components separately and adds them to obtain a reference total irradiance measurement. A calibration value is calculated based on the response near 45 solar zenith angle on clear days. In the shade/ unshade technique, a pyranometer is alternately shaded and unshaded on clear days near 45 solar zenith angle, and the difference in signal is compared to the direct component incident on the sensor as measured by a reference pyrheliometer. A third technique is to fit the pyranometer with a field-limiting aperture with the same field-of-view as the direct reference pyrheliometer and mount the pyranometer to track the sun. Forgan [1996] introduced a novel technique whereby pyranometers may be calibrated by switching shaded and unshaded pyranometers that measure diffuse and total irradiance, respectively, over the course of a few clear days using simultaneous measurements from a reference pyrheliometer. One or more of these techniques were used to calibrate the pyranometers in this study or to calibrate pyranometers that served as references for the pyranometers in this study. [6] Large negative offsets using single-black detector pyranometers can range between 10 and 20 W/m 2 in a clean, clear-sky situation with a true diffuse irradiance of about 100 W/m 2 [Bush et al., 2000; Haeffelin et al., 2001]. A pyranometer designed with a black and white detector that minimizes this offset can be used instead of single black detectors, but the ultimate solution for measuring diffuse irradiance remains elusive. There is no absolute standard for diffuse horizontal irradiance measurements or total horizontal irradiance measurements as exists for the direct normal irradiance. [7] As a first step toward establishing a working reference, we conducted an intensive observation period (IOP) in the last week of September and the first week of October 2001 to compare measurements of diffuse horizontal irradiance using pyranometers from most of the commercial vendors plus four prototypes. The goal was to determine whether there exists a consensus among these using current independent calibration practices. 2. Experimental Details [8] The participants included members of the ARM science team and BSRN communities [Ohmura et al., 1998] and pyranometer manufacturers. Table 1 is a list of the instruments, denoting whether they were ventilated, and the calibration method used. All pyranometers were mounted on Sci-Tec model 2AP two-axis trackers. Solar tracking during the experiment was sufficient so that we ascribe minimal uncertainty to tracking error. The instruments were all mounted with the surface of the detector horizontal and about 20 cm above the trackers pyranometer

3 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-3 shelves according to instructions in the 2AP manual. All were shaded by 5.1 cm spheres at about 51 cm from the center of the detector. While the shading was sufficient to block the direct beam for every instrument, the detector sizes vary, leading to differences in the amount of radiation received from the solar aureole. Calculated differences of diffuse based on estimated aerosol properties amounted to as much as 2 W/m 2. [9] On one of the trackers we also measured downwelling infrared radiation using an Eppley PIR pyrgeometer that was shaded from direct beam irradiance. These measurements are used to correct some of the offsets using one of the procedures proposed by Dutton et al. [2001]. [10] We expected to compare measurements on both clear and cloudy days; however, during the two weeks of the experiment bright sunshine was detectable for more than 95% of the daylight hours, and during much of this time the skies were completely cloudless. There were a few totally overcast hours, fortunately, to assess the pyranometers performance for these conditions relative to clear skies. The preponderance of clear skies meant that there were many occasions to model the diffuse and compare to measurements. Direct beam absolute cavity radiometer measurements were obtained using the Pacific Northwest National Laboratory s TMI radiometer that had been calibrated to the World Radiometric Reference the previous year at the World Radiation Center in Davos, Switzerland [Rüedi et al., 2000]. [11] To determine instantaneous offsets during the daylight hours, capping experiments were performed as described by Dutton et al. [2001]. The goal was to quickly cap the pyranometer to block visible light and measure the response of the instrument caused by the dome s cooler temperature relative to the case temperature of the detector. This requires that the detector response is rapid relative to the time required for the dome to change temperature after the cap is in position. The capping was performed four times with enough time between cappings to allow the pyranometers to recover to near their steady state. The instruments were capped while measuring diffuse in the midafternoon on a clear day because we find that the largest offsets occur for these conditions. 3. Results 3.1. Diffuse Irradiance Comparisons [12] After examining the total data set we found the most consistent behavior among five commercial pyranometers, moderate consistency among three of the five remaining commercial pyranometers, and less consistency among the prototypes. Our focus was initially on clear-day measurements since those were the most likely in error because of thermal offsets and asymmetric skylight distributions falling on instruments with imperfect cosine responses. Figure 1 contains data from the clear day of 28 September 2001 (day of year 271). The data plotted are for solar elevations greater than -10. Time is local standard time; for the Southern Great Plains (SGP) site this means that solar noon occurs about 20 min after local noon during the IOP. The psp-mh data are corrected according to the procedures outlined by Haeffelin et al. [2001]. The four others have no corrections. The three Kipp & Zonen (cmxx) pyranometers in this figure have very high volume ventilation relative to the other pyranometers in this experiment; the cm22-rp slightly heats the vented air before it passes over the dome. The psp-mh and the 8-48 have modest ventilation to keep dew from forming on the outer dome. The 1-min difference of each measurement from the mean of the five is plotted in the bottom of Figure 1 (left-hand ordinate). The standard deviations among the five are plotted as a function of time according to the ordinate label on the right. The range in standard deviation for this day was W/m 2 with an average deviation of 0.63 W/m 2. [13] Note that for each of the three days of data presented in the first nine figures we assume that the average of the five pyranometers listed in the first of each of the three figures for the day in question is our standard for comparison based on their consistent pattern of agreement. In fact, since there is no absolute standard for diffuse measurements, we cannot be certain of the absolute accuracy of this average. All plots are on the same scale to facilitate comparison. [14] In Figure 2 the psp has been corrected using one of the methods outlined by Dutton et al. [2001] where the offset is determined using nighttime pyranometer data (sun lower than -10 ) regressed against the net PIR signal with the intercept forced through zero (hereafter, this is denoted the net-pir method). This was also used for the cm21, which had a lower volume ventilation system than the Kipp & Zonen pyranometers in Figure 1. The Meteorological Services of Canada (MSC) normal correction algorithm produces a correction that did not work as well; it uses a time-based interpolation between the average presunrise and postsunset thermal offset. The MSC network does not include pyrgeometers at most of its sites, which precludes the routine use of the net-pir method. The cimel, eko, and schenk data were used without correction. The psp and cm21 were ventilated to prevent dew formation, but the cimel, eko, and schenk were not. These five measurements are plotted in the top of Figure 2. The mean of the five measurements in Figure 1 is included for comparison. The bottom of Figure 2 is a plot of the differences between the mean in Figure 1 and each of the pyranometers of Figure 2. [15] The four prototypes that were included in the experiment (eq08, new_ep_bw, yes, and tsbr) are plotted in the top of Figure 3 along with the mean from Figure 1, and the differences from this mean are plotted in the bottom of Figure 3. It appears that some improvement could be achieved with a better calibration and/or offset correction, although it is not clear that a simple adjustment is possible for every prototype. [16] We observed that the cimel and eq08 were more sensitive to wind gusts than other instruments. Compare the fluctuations in the bottoms of Figures 2 and 3 with the other pyranometers fluctuations in those figures. When the wind was calm we were less likely to notice large amplitude fluctuations in the data. [17] Figures 4, 5, and 6 are for another clear day (2 October 2001, day of year 275) presented in the same way as Figures 1 3 above. This is a lower aerosol day with smaller diffuse. The average standard deviation among the five pyranometers in Figure 4 for the day was 0.73 with a range of W/m 2. The spread among the pyranometers in Figure 5 is slightly greater than it was for day 271 in Figure 2, especially for the eko. A similar pattern of disagreement among the prototypes appears in Figure 6 as

4 AAC 9-4 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Figure 1. Diffuse horizontal irradiance versus time for 28 September 2001 (top). This plot includes five measurements that were the most consistent for both clear and overcast conditions throughout the IOP. The bottom plot contains deviations from the mean of the group (left-hand ordinate) and the standard deviation among the group of five (right-hand ordinate). The standard deviations are generally less than 1 W/m 2. we saw for day 271 in Figure 3 with the exception of the tsbr, which is closer to agreement with our comparison standard than it was in Figure 3. [18] In the morning of 5 October 2001 (day of year 278) we had a few hours of opaque cloud cover, the only total cloud cover of the IOP. Figures 7 9 are plots for the same sets of instruments as presented in the previous two sets of figures for this day that is overcast until (before the red vertical line). After the skies were partly cloudy until sunset. In Figures 7 9 the fractional deviation from

5 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-5 Figure 2. A plot similar to Figure 1 for the second most consistent group of diffuse radiometers along with the mean from Figure 1. The deviations plotted on the bottom of this figure are those from the mean of Figure 1. The deviations are about twice that of Figure 1. the mean of the same five instruments is plotted in the bottom of the plots rather than the absolute difference in W/m 2. The vertical blue lines delimit the time of day when the irradiance exceeds approximately 50 W/m 2. Lower values than this in the early morning and late afternoon show large scatter in fractional values because of the low irradiance. There is a fairly tight grouping in Figure 7 of the five pyranometers that were the most consistent in earlier

6 AAC 9-6 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Figure 3. A plot similar to Figure 2 for the four prototypes included in the study. Note that the deviations in the bottom of the figure are considerably larger than for Figures 1 and 2. figures. After , the sun broke through the clouds intermittently for the rest of the day. The scatter in irradiance among the instruments is larger for these conditions, but the absolute value of the diffuse irradiance is much higher as well, such that the fractional scatter is compara- ble. The next most consistent group of pyranometers shown in Figure 8 shows a larger spread for the total overcast conditions before than they did for clear skies and a comparable grouping for partly cloudy skies. In the bottom of Figure 9 the prototype differences from the mean in

7 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-7 Figure 4. A plot similar to Figure 1 for 2 October This day had a smaller aerosol burden and therefore lower diffuse. The results are similar to those in Figure 1. Figure 7 show large inconsistencies as they did for the clear days in Figures 3 and Zero Offsets [19] The left-hand sides of the 13 parts of Figure 10 contain plots of uncorrected pyranometer readings at night (sun elevation less than 10 ) versus the net PIR signal for 13 of the 14 pyranometers. The solid line is the least squares fit to the data with the intercept forced through zero; the dashed line is the least squares fit to the data with the intercept unrestricted. The uncorrected psp, psp-mh, and cm21 have the largest offsets; the psp and cm21 were corrected in Figures 2, 5, and 8 using the net-pir method (the solid line fits); and the psp-mh was corrected in Figures

8 AAC 9-8 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Figure 5. Same day as Figure 4 for the second group of instruments as in Figure 2. The deviations are somewhat larger than in Figure 2. 1, 2, and 3 using the measured temperature difference between the inner glass dome and the case according to the procedure given by Haeffelin et al. [2001]. The 8 48, cm11, cm22, cimel, and eko have smaller dependencies on the net PIR signal and were not corrected for offset. The cm22-rp, eq08, new_ep_bw, schenk, and yes offsets do not depend on net PIR, or are almost zero with the exception of the new_- epp_bw, which has a positive offset of about 4 W/m 2.

9 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-9 Figure 6. Same day as Figure 5 for the prototypes. The deviations are similar to Figure 3 with the exception of the tsbr, which is now higher than the mean of Figure 4. [20] The right-hand side of the 13 parts of Figure 10 shows the uncorrected results of an experiment performed when the instruments, measuring diffuse, were capped in the early afternoon of a clear day (29 September 2001, day of year 272). Samples were written to the file every two seconds. The capping was repeated four times. The net infrared signal measured by the PIR was a 146 W/m 2 during this time. If we assume that the net-pir method

10 AAC 9-10 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Figure 7. Diffuse horizontal irradiance versus time for the overcast and then partly cloudy day of 5 October This plot includes five measurements that were the most consistent group for both clear and overcast conditions throughout the IOP. The bottom plot contains deviations from the mean of the group as a fraction of the diffuse irradiance. The deviations are generally less than applies to the instruments that show some dependence on net PIR signal as discussed earlier, we can estimate the expected offset at time of capping. However, there has been no previous demonstration that the net-pir method of offset correction applies to any of these instruments except the moderately ventilated Eppley PSP. [21] The 8 48 goes slightly negative upon capping, which is consistent with an expected 1 to 2 W/m 2

11 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-11 Figure 8. The second most consistent group of the comparison on the cloudy day of 5 October The deviations are about twice as large at those of Figure 7. offset from the nighttime estimate and the slower time response of this instrument. Both uncorrected psp and pspmh show the expected large capped offset from the netpir estimate. The cm21 estimated offset based on the net-pir method is about 6 W/m 2, which is slightly smaller than the 7 8 W/m 2 offset from capping. The cm11, cm22, and cm22-rp capped offsets are small and consistent with the nighttime estimates. [22] The eq08 and cimel were sensitive to gusty winds, as mentioned earlier, and both show relatively unstable behav-

12 AAC 9-12 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Figure 9. The deviations among the prototypes for 5 October 2001 are much larger than in Figures 7 and 8. ior if we compare them to the stability of the other instruments in the capping experiment at full diffuse signal. The eq08 according to the left-hand side of Figure 10 does not depend on the net PIR signal, but the amplitude of the noise in the left-hand figure is high. The amplitude of the noise in the uncapped and capped signal on the right-hand side of this figure is consistent with the noisy nighttime signal. For the net PIR signal measured during this time we might expect a cimel offset of 2 to 3 W/m 2, which again is plausible given the noise in the uncapped signal. The wind effects on the signal make this a difficult confirmation without many more capping tests and averaging.

13 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-13 The left-hand sides of these plots include only data with the sun at least 10 below the horizon. The plots are the offsets of the diffuse irradiance as a function of the net infrared measurements made by the Eppley PIR pyrgeometer. These plots are explained by Dutton et al. [2001]. The dashed lines are linear fits through the points, and the solid lines are fits through the points forced to zero diffuse at zero net infrared irradiance. Ideally, these plots would show no dependence on net infrared as seen for the schenk and the yes. The uncorrected psp and psp-mh have the largest offsets, and the others are in between. Some of these offsets are corrected in the first nine figures above using the Dutton et al. [2001] correction procedure, while others have not been corrected. The right-hand sides of these plots demonstrate a procedure for estimating the zero offset instantaneously during the daytime. The diffuse radiometer is capped quickly and the response is measured before the dome has time to change temperature significantly. This assumes that the detector response is fast relative to the time it takes for the dome temperature to change. All except two of these pyranometers (the eko and the new_ep_bw) show capping results that are consistent with the offsets predicted from the nighttime measurements given the net infrared at the time of capping. [23] The eko expected offset was about 2W/m 2, but the offset as judged by the lowest dip is 2 3 times this. The new_ep_bw was expected to have a positive offset judging from the data at night in the left-hand side of Figure 10, but goes negative by a few W/m 2 in the right-hand side. The schenk is barely negative, which is consistent with nighttime data, and the yes is zero, consistent with nighttime results. 4. Summary and Discussion [24] Fourteen measurements of diffuse horizontal irradiance were made simultaneously over a two-week period in September and October Tracking to keep the instruments shaded was very good, and there was a preponderance of clear, cloud-free weather. Five of the measurements were generally within 1 2 W/m 2 of their mean for both clear and cloudy conditions. Three other measurements were only slightly less stable with most measurements within 2 3 W/m 2 of this same mean. The four prototype instruments showed less agreement with the most consistent group than most of the commercial instruments. [25] Figure 11 summarizes these results. The dark bars represent the root-mean square (RMS) differences from the mean of the five most consistent measurements. Since these RMS differences include the bias (light bars) from the mean of the same five measurements, they overestimate the standard deviation; however, they provide a conservative estimate of reproducibility. The positive bias of the schenk is notable because it has a very large receiver, and since we used the same shading apparatus for all pyranometers, it received more radiation from the solar aureole than most. Had we corrected for this effect, it may have lowered this bias and reduced the RMS difference. [26] There is a suggestion from Figure 10 that there may be notable benefits associated with high-volume ventilation.

14 AAC 9-14 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON

15 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-15

16 AAC 9-16 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON

17 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-17

18 AAC 9-18 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON

19 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-19

20 AAC 9-20 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON Figure 11. Plot summary for the two-week IOP. Dark bars are root-mean square differences from the mean of the five most consistent results. Light bars are biases with respect to the mean of the same five measurements. The cm11 and cm22 used the high volume CV2 ventilation system of Kipp & Zonen to minimize the temperature difference between case and dome. These single-black detectors would be expected to have large thermal offsets but are among the group showing little nighttime offset and capping results that are consistent with their small nighttime offsets. Moreover, their small offsets indicate excellent predictability using the net-pir method. The cm21 used a different ventilator that moved much less air over the dome than the CV2 does. The cm22-rp used the CV2 ventilation system but modified the air stream by heating the air just before it exited the ventilator. This appears to introduce an offset in the net-pir relationship so that the readings are slightly negative for large negative values of the net infrared (clearest skies) and slightly positive for small negative values of the net infrared (cloudy skies) with the difference from zero generally less than 1 W/m 2.Of course, these issues should be studied further with higher and lower ventilation for all pyranometers in order to investigate the possible benefits. [27] There are issues regarding calibration that need to be reexamined. The current procedure in ARM produces calibrations that result in higher diffuse irradiances than are measured by the most consistent group of five. The 8 48 and PSP that showed consistency with this group used Eppley factory calibrations rather than the ARM calibrations. ARM calibrations would give readings about 3 4 W/m 2 higher in the clear-sky examples shown. [28] There are geometry differences in the radiometer detectors, as noted above for the schenk. Since all were shaded similarly with the same size blocker at a fixed distance, the larger detectors receive more diffuse radiation from the penumbra than the smaller detectors. The largest differences from this effect will occur in clear, hazy conditions. Figure 12 is a plot of the differences in diffuse irradiance received for a range of clear-day continental aerosol conditions thought typical of those encountered during this IOP based on information given by Major [1994]. There are four sizes of receivers with diameters of about 32, 20, 11, and 4 mm. Calculations suggest that the largest differences in aureole diffuse are no more than about 2W/m 2 as shown in Figure 12. [29] In this study the Campbell Scientific data loggers measure voltages, on their smallest scale, accurate to about 5 mvolts or about 0.5 W/m 2. This is near the level at which we are measuring differences, and greater care will be required, especially with regard to understanding possible electronic offsets, in any follow-on experiment. [30] Based on this study and work to follow where we will refine the procedures used in this initial study it appears that we should be able to establish a set of instruments that we can maintain as a working diffuse standard group for the ARM and BSRN communities. However, the question of how close this standard is to the true absolute value will remain elusive until an absolute diffuse radiometer is developed.

21 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON AAC 9-21 Figure 12. Plotted are the estimates of the differences in diffuse irradiances from the aureole based on differences in the sizes of the receivers of the radiometers as a function of diffuse irradiance produced by a range of continental aerosol conditions. The four sizes are approximately 32, 20, 11, and 4 mm diameters with the largest detector receiving the most aureole irradiance. The three curves are for the differences between the largest and each of the other sized detectors. [31] Acknowledgments. The efforts of the ARM staff at the SGP site, especially Craig Webb, are gratefully acknowledged. Brett Bush, Sabrina Leitner, David Marsden, and Francisco Valero provided their tsbr data for inclusion in this paper, and their efforts are appreciated. Chuck Long and an anonymous referee s comments added needed clarity to the paper. Baseline Surface Radiation Network sponsorship of this effort is appreciated. This research was supported by the Biological and Environmental Program (BER), U. S. Department of Energy, through grant DE-FG02-90ER References Bush, B. C., F. P. J. Valero, A. S. Simpson, and L. Bignone, Characterization of thermal effects in pyranometers: A data correction algorithm for improved measurement of surface insolation, J. Atmos. Ocean. Tech., 17, , Dutton, E. G., J. J. Michalsky, T. Stoffel, B. W. Forgan, J. Hickey, D. W. Nelson, T. L. Alberta, and I. Reda, Measurement of broadband diffuse solar irradiance using current commercial instrumentation with a correction for thermal offset errors, J. Atmos. Ocean. Tech., 18, , Forgan, B., A new method for calibrating reference and field pyranometers, J. Atmos. Ocean. Tech, 13, , Haeffelin, M., S. Kato, A. M. Smith, C. K. Rutledge, T. P. Charlock, and J. R. Mahan, Determination of the thermal offset of the Eppley precision spectral pyranometer, Appl. Opt., 40, , Halthore, R. N., and S. E. Schwartz, Comparison of model-estimated and measured diffuse downward irradiance at surface in cloud-free skies, J. Geophys. Res., 105, 20,165 20,177, Halthore, R. N., S. E. Schwartz, J. J. Michalsky, G. P. Anderson, R. A. Ferrare, B. N. Holben, and H. M. Ten Brink, Comparison of model estimated and measured direct-normal solar irradiance, J. Geophys. Res., 102, 29,991 30,002, Kato, S., T. P. Ackerman, E. E. Clothiaux, J. H. Mather, G. G. Mace, M. L. Wesely, F. Murcray, and J. Michalsky, Uncertainties in modeled and measured clear-sky surface shortwave irradiances, J. Geophys. Res., 102, 25,881 25,898, Major, G. V., Circumsolar correction for pyrheliometers and diffusometers, WMO/TD-No.635, 42 pp., World Meteorol. Org., Geneva, Switzerland, Ohmura, A., et al., Baseline surface radiation network (BSRN/WCRP): New precision radiometry for climate research, Bull. Am. Meteorol. Soc., 79, , Rüedi, I., C. Fröhlich, W. Schmutz, C. Wehrli, Report on the International Pyrheliometer Comparisons The maintenance and dissemination of the World Radiometric Reference, IOM Rep. 74 (WMO/TD-No. 1028), pp , World Meteorol. Org., Geneva, Switzerland, Valero, F., P. J., and B. C. Bush, Measured and calculated clear-sky solar radiative fluxes during the Subsonic Aircraft Contrail and Cloud Effects Special Study (SUCCESS), J. Geophys. Res., 104, 27,387 27,398, Zender, C. S., B. Bush, S. K. Pope, A. Bucholtz, W. D. Collins, J. T. Kiehl, F. P. J. Valero, and J. Vitko Jr., Atmospheric absorption during the Atmospheric Radiation Measurement (ARM) Enhanced Shortwave Experiment (ARESE), J. Geophys. Res., 102, 29,901 29,915, R. Dolce, Kipp & Zonen, Inc., 125 Wilbur Place, Bohemia, NY 11716, USA. (robert.dolce@kippzonen.com) E. G. Dutton, CMDL/NOAA, 325 Broadway, Boulder, CO , USA. (edutton@cmdl.noaa.gov) M. Haeffelin, Laboratorie de Meteorologie Dynamique, Ecole Polytechnique, Palaiseau Cedex, France. (haeffelin@lmd.polytechnique. fr) J. R. Hickey, The Eppley Laboratory, Inc., 12 Sheffield Ave., Box 419, Newport, RI 02840, USA. ( jhickeyeplab@ids.net) W. Jeffries, Yankee Environmental Systems, Inc., 101 Industrial Blvd., Turners Falls, MA 01376, USA. (wqj@yesinc.com) A. Los, Kipp & Zonen, Inc., Röntgenweg 1, 2624 BD Delft, P. O. Box 507, 2600 AM Delft, Netherlands. (alexander.los@kippzonen.com) G. Major, Hungarian Meteorological Society, P. O. Box 433, H-1371 Budapest, Hungary. (h10830maj@helka.iif.hu)

22 AAC 9-22 MICHALSKY ET AL.: DIFFUSE HORIZONTAL IRRADIANCE COMPARISON D. Mathias, Carter-Scott Design, 16 Wilson Ave., Brunswick, Victoria, 3056 Australia. L. J. B. McArthur, Meteorological Service of Canada, Downs-view, Ontario, M3H5T4 Canada. J. J. Michalsky and J. A. Schlemmer, Atmospheric Sciences Research Center, State University of New York, Albany, 251 Fuller Rd., Albany, NY 12203, USA. ( jim@asrc.cestm.albany.edu; joe@asrc.cestm.albany. edu) R. Philipona, Physikalisch-Meteorologisches Observatorium Davos, World Radiation Center, Dorfstrasse 33, CH-7260 Davos Dorf, Switzerland. (rphilipona@pmodwrc.ch) D. W. Slater, Pacific Northwest National Laboratory, P. O. Box 999, Richland, Washington 99352, USA. (donald.slater@pnl.gov) I. Reda and T. Stoffel, National Renewable Energy Laboratory, 1617 Cole Blvd., Golden, CO , USA. (ibrahim_reda@nrel.gov; thomas_stoffel@nrel.gov)